14 November, 2019
Model Answers and Grades are online!
Feedback will follow soon.
The Midterm Exam will be discussed on Nov 21, Nov 28 (and possibly Dec 5) from 13.15 - 14.00 hours.
Erratum in the exam? R version related…
Somehow the encrypted student names where not the same as you may have expeced. This was related to the sample() function with which the key20191030 data.frame was generated.
Apparantly, even if one sets a seed (using set.seed()) the default random generator works different for the R versions < 3.6.0, compared to R versions >= 3.6.0.
In R versions 3.6.0 the Random Number Generator got updated with a new sampling strategy, see https://bugs.r-project.org/bugzilla/show_bug.cgi?id=17494
This presentation is made in R version 3.6.1. While working in a R version >= 3.6.0, by default we would have
set.seed(191114) sample(x = 1:10, size = 2)
To mimic the default random number generators for R versions < 3.6.0, we would need the code
RNGkind(sample.kind = "Round") # back to default R versions < 3.6.0 set.seed(191114) sample(x = 1:10, size = 2)
To use the default sample.kind again:
RNGkind(sample.kind = "default")
Goal:
Obtain a good idea of the sampling distribution of an estimator/statistic under a certain hypothesis (e.g. nullhypothesis)
Monte Carlo studies:
the true distribution of each data point is assumed known and used to set up a simulation s.t. we can obtain a Monte Carlo sampling distribution of any estimator of interest.
Goal:
Obtain a good idea of the sampling distribution of an estimator/statistic under a certain hypothesis (e.g. nullhypothesis)
Permutation context:
there is not necessarily a known distribution for each data point. We use (random) permutations in our data to set up a Permutation sampling distribution, assumed to be a conditional sampling distribution in accordance with the hypothesis.